This is precisely because of free rider problems. Ultimately we only have one world to inhabit and only one chance to get it right.
This is a somewhat different condition from normal market conditions where there is usually a chance to restart. Also, unlike a normal market, participants are all sovereign nations, with no real way to enforce behavior.
If the actions of any set of actors triggered abandonment then we would be back at square one, having run down the clock a bit more. This is precisely what we want to avoid.
Early game theory such as the prisoner dilemma focussed on zero-sum game, one-off situations. They are hugely influential in decision analysis, but additional studies in the field have also looked at recurring events, where actors gradually can gravitate towards more collaborative behavior.
The experimental data show that cooperation is affected by infinite repetition and is more likely to arise when it can be supported in equilibrium. However, the fact that cooperation can be supported in equilibrium does not imply that most subjects will cooperate. High cooperation rates will emerge only when the parameters of the repeated game are such that cooperation is very robust to strategic uncertainty.
In this case far better to name and shame the shirkers on the next round. Reality is we are still learning how to collaborate on this problem, with many actors having differing priorities but still mostly being faced with negative outcomes if we get it wrong enough (for example, the Arabian Gulf states may want to export gas as long as possible but their geographic area is also projected to start hitting human wet bulb heat tolerance limits).
As opposed to Kyoto (1992) one of the new approaches since Paris (2015) has been ratchets or NDC (nationally determined contributions). Rather than committing all at once to "fixing everything" like they did before, nations commit to self-chosen goals to limit emissions, which are then expected to strengthen every 5 years. This gives them opportunities to get back in shape, without just throwing up their hands (as Canada did from as soon as it signed Kyoto, for example). And since these are self-chosen, they are hopefully more likely to be followed up on, building trust amongst the actors.
Note: as remarked upon in the comments, the most famous game theory problem, the prisoner dilemma, is not a zero sum game. However, as its maximum utility from any participant's point of view is when they "screw over" the other person when the other person is collaborating it tends to not generate much collaboration when only played once, especially in conditions where trust is lacking. I don't pretend to be an expert in game theory or its terminology, so I admit to stretching this analogy a bit.